Carnot execution simulation application

[danielSanchezQ]

We need to be able to simulate execution times based on different parameters. The simulation app should not actually execute consensus, but black-box it over custom configurable execution times.

Current base effort is done at #90. But this is just a start point and may/should change. It also have things that need to be done.

Tasks:

• [x] Configuration (taken from IO)

  • [x] Networking cost should be configurable (fixed or ranged).
  • [x] Let user configure node steps

• [x] Regions should be configurable as well

• [x] Overlay is just implemented for Flat, and takes no settings. We should implement a binary tree overlay distribution that is configurable over the size of the committees.

  • [ ] Tree overlay configuration should be taken from IO.

• [ ] Carnot execution with the runner must be exhaustively tested. We need to make sure output simulation data it is what should be

  • [ ] Use different config files to create multiple testcases for runner
  • [ ] Add network drop unit tests

• [ ] Make network wait maximum delay if a message gets dropped. https://github.com/logos-co/nomos-research/pull/90#discussion_r1129216403

• [ ] Implement Carnot node

  • [x] #168
  • [x] #169

Chores

• [x] CommitteeId and NodeId are just type aliases, it lead to a bug where we were mixing them. Make them actual wrapper types to avoid this. #104

Apps

• [ ] CLI
• [ ] WASM web app

[danielSanchezQ]

This is the initial list of things to do. @al8n @bacv, please edit the list as completed or add new items as needed. The relate the PRs, or create tasks in the board for each item you take. 🚀

[bacv]

Notes regarding next tasks:

• [x] Warding - Analyze simulation state and implement the stopping of simulation.
  • [x] The difference minimum view and and maximum view in the individual nodes need to be tracked. #108
  • [x] Track stalled nodes (e.g incoming queue is empty, the node doesn't write to other queues). #109
• [x] Output processor - Gather states from every node during each step. Data needs to be per view. #111
• [x] Runners
  • [x] Synchronous runner - track the state and decide when to dump the data #111
  • [x] Asynchronous runner - track the state and decide when to dump the data #111
  • [x] A cache that stores the view number and the vector of nodeids and state. #119
    • [x] Update method - updates node state for that view #119
    • [x] Integrate cache withing the runners #119
• [x] Mock Carnot node - every step will increment the view count. #114
• [x] Carnot node settings:
  • [x] Configurable network behavior. #110
  • [x] Layout per view, accessible withing the node. #114
• [ ] Overlay trait needs to accept seed parameter when the layout is generated. https://github.com/logos-co/nomos-node/pull/122#discussion_r1187134287
• [ ] Investigate if ViewId + ViewOverlay pairs need to be shared state. Nodes should recreate an overlay from seed. (Mentioned in https://github.com/logos-co/nomos-research/pull/114#discussion_r1159808313)